Disentangling Superpositions: Interpretable Brain Encoding Model with Sparse Concept Atoms | Read Paper on Bytez